AI and the challenge of listening to public opinion.

Artificial intelligence is opening up many new possibilities in public opinion research and policy making. However, despite the benefits of speed and data processing capabilities, AI cannot replace the process of dialogue, listening, and truly understanding people's opinions.

AI opens up new possibilities in public opinion research.

For many years, sociological surveys, in-depth interviews, and focus groups have remained the primary methods for governments and organizations to gauge public opinion. The rapid development of artificial intelligence is bringing about significant changes to this field.

In 2025, researchers at Stanford University published test results showing that AI could simulate survey responses from over 1,000 individuals based on in-depth interview data, with an accuracy of up to 85%. This result demonstrates the potential of AI to become a useful tool in policymaking, particularly in predicting social reactions to new proposals.

A major advantage of AI is its ability to process massive amounts of data in a short time, detecting trends, patterns, or issues that humans might miss. Compared to traditional surveys, which are time-consuming and costly, AI can help managers quickly obtain initial assessments of public opinion.

However, the use of AI in public opinion polling has also sparked controversy. Recently, an article by the media outlet Axios cited "survey results" from the AI ​​startup Aaru regarding public trust in doctors and nurses. It was later discovered that all the data was generated from computer simulations, with no real people participating in the survey. This incident raised concerns about the risk of misleading the public, as well as ethical issues related to "silicon sampling"—a practice using large language models to create simulated responses from "digital actors."

Direct dialogue and listening are irreplaceable.

AI can help analyze public opinion faster, but it cannot replace listening to and understanding the people. Photo: Lim Li Ting/Today

While AI can support fast and efficient data analysis, this technology still struggles to fully reflect the complexity of social life. Individual perspectives are shaped not only by rational information but also by emotions, experiences, and specific circumstances.

One might agree with the policy's objectives but have concerns about its implementation. Initial opinions may be supportive, but deeper discussions reveal various other worries. These are aspects that AI models struggle to fully capture based solely on aggregated data.

In Singapore, the practice of listening to citizens is still carried out through various direct forms such as surveys, community dialogues, and group discussions. The Singapore Government's Response Unit (REACH) regularly organizes in-person and online consultation programs on social issues such as the cost of living, employment, and welfare. The Forward SG program even attracted over 200,000 citizens to contribute their opinions through various channels.

These activities are not only aimed at gathering responses, but also at helping authorities understand the underlying reasons, sentiments, and concerns behind each opinion. This is a core element in building social consensus and strengthening public trust in public policy.

AI can help quickly identify trends in public opinion or recurring issues in social responses. However, this technology cannot completely replace face-to-face interaction – where people can express their thoughts, feelings, and even things they are hesitant to say.

Use AI responsibly, putting people at the center.

One of the biggest risks today is the trend of over-reliance on AI. As models become more sophisticated, many organizations may become caught up in the convenience of the technology and gradually cut back on field surveys, interviews, or direct dialogue with citizens.

This risks transforming AI from a supportive tool into a "shortcut" that replaces the process of listening to society. Meanwhile, many important issues only emerge through follow-up questions, in-depth discussions, or psychological expressions that aggregated data may not fully reflect.

Furthermore, while AI-generated results may appear objective and scientific, they can still be affected by incomplete, outdated, or biased input data. Some social groups may be overrepresented, while marginalized or less prominent voices in the digital space may be overlooked.

When AI is used to build “synthetic audiences” to represent community thinking, many people may feel they aren’t truly being heard. If an individual is told that “the model predicted what you might think,” instead of being asked directly, social trust is bound to suffer.

Therefore, the issue is not whether or not to use AI in public opinion research, but how to use it correctly and responsibly. AI should be seen as a supporting tool to improve the efficiency of analysis and forecasting, while humans must still retain a central role in the policy-making process.

To achieve this, it is necessary to ensure transparency in data collection and processing; maintain forms of direct consultation; and consider public trust as a core element in the design and application of technology.

AI can help society better understand the dynamics of public opinion. But ultimately, technology is only truly meaningful when it contributes to helping people listen to each other better, not when it replaces listening.

Source: https://daibieunhandan.vn/ai-va-bai-toan-lang-nghe-du-luan-xa-hoi-10418843.html